Most AI agent pilots never reach production. The model works; what fails is everything around it — the security review, the data-residency question, the change-control board, the auditor's first request. ATX Agent Hub by ATX Institute is built from that list backwards: an AI agent platform that a bank, an insurer or a public body can install in its own cloud, connect to its identity provider and put through the same governance as any other production system.
ATX Institute was founded by two Business Intelligence and AI experts with twenty years of building enterprise analytics platforms in major European banks — thousands of users, petabytes of data, and the audit, risk and security functions that come with them. ATX Agent Hub encodes what those functions ask for, so that an AI agent can pass the review instead of stalling in it.
The Enterprise plan deploys ATX Agent Hub into your own Google Cloud or Microsoft Azure subscription with Terraform included, or on-premise and air-gapped. Your documents, embeddings, run history and credentials stay in your tenancy; models are reached through your own Azure AI Foundry or Vertex AI endpoints.
People sign in through your identity provider — SSO with SAML, LDAP and SCIM provisioning on the Enterprise plan — and every permission is a role: owner, editor, runner, approver and auditor per project, with platform administrators and auditors above them. Projects are hard isolation boundaries: workflows, knowledge bases, credentials, manuals and run history belong to exactly one project.
Every workflow version is immutable. Promotion runs DEV → TEST → PROD, and promotion to production needs an independent approver — the requester cannot approve their own change. Test suites can gate the promotion, and every run records its inputs, outputs, the evidence retrieved and the reasoning path across each step.
AI Guardian, Fact Checking, Intent Check and Domain Constraint nodes evaluate output before it is delivered. Human-in-the-loop nodes stop a workflow until a named person approves. Token use and cost are tracked per run, workflow, project, user and model, with soft caps that warn and hard caps that block.
We answer security questionnaires and architecture reviews directly, with the people who built the platform. The pages below cover the ground most reviews start with.
Deployment in your own cloud or data centre, single sign-on and provisioning, role-based access with separation of duties, immutable versions with independent approval before production, output guardrails, human approval for consequential actions, full trace logs, cost limits and testing. ATX Agent Hub provides each of these.
Yes. The Enterprise plan is self-hosted on Google Cloud or Microsoft Azure, with Terraform included, and can also be deployed on-premise or air-gapped.
Yes. SSO with SAML, LDAP and SCIM provisioning is part of the Enterprise plan, and the end-user portal is SSO-gated.
ATX Institute, an independent European software company founded by two Business Intelligence and AI experts with twenty years of experience building enterprise platforms in major European banks.
This site needs JavaScript for the full experience. Every page above is reachable without it.